NovaVoice
NovaVoice is a desktop voice assistant for macOS, Windows and Linux. It dictates into any app, reformats text on command, executes approved actions in connected apps, and answers on-screen questions. A permanent free plan is available.
What is NovaVoice?
NovaVoice, which the desktop app itself calls Nova, is a voice layer that sits on top of the operating system rather than inside a single application. It is published by Xware VC LLP, which gives Astana, Kazakhstan as its country, and it runs on macOS 12 and later, Windows 10 and 11, and Linux (Ubuntu, Debian, Fedora). A global hotkey activates it, and whatever it produces lands in the text field that currently has focus, so no per-application integration is required.
The product is organised around four modes. Dictation Mode turns speech into punctuated, formatted text in any input field; the homepage contrasts 45 words per minute of typing with more than 200 spoken, while the developer page uses more modest figures, 130 spoken against 40 typed. Formatting Mode rewrites what you have just said into a chosen style, whether a one-off spoken instruction, a saved style, or a rule attached to a particular application window, so that a pull request receives structured prose while a chat window receives something looser. Agent Mode carries voice commands into connected applications: the product pages list Gmail, Todoist, Google Calendar, Telegram, WhatsApp, Spotify, X and Hacker News. Every action is displayed for explicit approval before it runs. Assistant Mode answers spoken questions, including questions about what is currently on screen, without opening a browser.
A Terms Dictionary underpins all four modes: you register project names, technical jargon, contact addresses, postal addresses and loyalty numbers so that they are transcribed correctly and can be inserted with a single command.
Beyond the individual product, a Team tier adds shared styles, a shared dictionary, priority support and centralised billing, while an enterprise option deploys the processing server inside the customer's own data centre, private cloud or air-gapped network, so that audio never leaves their infrastructure. NovaVoice is young, its domain having been registered in July 2025, but visibly active: a public changelog runs from version 1.8.9 to version 1.8.21 across January and February 2026.
What it does
- Dictate into any desktop application through a single global hotkey
- Reformat what you just dictated into any style, by voice
- Execute actions in connected apps, after approving each one
- Ask a built-in AI assistant about whatever is on your screen
- Insert saved names, addresses, numbers and jargon on command
- Apply a different formatting rule to each application window
- Share formatting styles and a common dictionary across a team
When to use NovaVoice / When not to
A quick filter to help you decide if NovaVoice is the right fit.
When to use NovaVoice
- Developers who write far more prose than code: commit messages, pull request descriptions, documentation and tickets
- Sales and customer-facing teams capturing CRM notes, emails and follow-ups immediately after a call
- Linux desktop users, who are rarely served by commercial dictation software
- IT teams standardising a single dictation tool across a mixed Mac, Windows and Linux fleet
- Legal, financial, healthcare and public-sector teams that need voice processing to stay inside their own network
When not to use NovaVoice
- Mobile-first users: there is no iOS or Android app, only desktop clients
- Developers wanting to embed dictation in their own product, since no public API or developer documentation exists
- Buyers who require GDPR documentation, a data processing agreement, a subprocessor list or published certifications
- Anyone hoping to dictate raw source code character by character, which the vendor explicitly rules out
- Teams needing deep native integrations with GitHub, Jira, Linear, Slack or Notion, which Agent Mode does not cover today
How to use NovaVoice
A typical end-to-end flow, from setup to results.
- Download the build for your platform from the downloads page: macOS for Apple Silicon, Windows 10 or 11, or a .deb package for Debian and Ubuntu
- Install the application and sign in to your NovaVoice account
- Learn the global hotkey, which is what wakes the assistant from any application
- Put the cursor in any text field, hold the hotkey and speak; punctuated text appears where you were typing
- Say a formatting instruction, or pick a saved style, to have that text rewritten in place
- Attach a default formatting rule to a given application window so the right style applies automatically
- Connect the accounts you want Agent Mode to reach, such as Gmail, Todoist or Google Calendar
- Issue a spoken command, then read the proposed action and approve or deny it
- Register recurring names, jargon, addresses and numbers in the Terms Dictionary
- For a team, move to the Team plan for shared styles, a shared dictionary and one invoice; for on-premise, have your IT team deploy the processing server
Pros & Cons
Pros
- A genuine Linux desktop client, which almost no commercial dictation tool offers
- One product across macOS, Windows and Linux, which suits mixed device fleets
- Works at operating-system level in any application, with no per-app setup
- Low entry cost: a permanent free plan, USD 10 per month for Standard, USD 8 per seat per month for Team
- Agent Mode requires explicit approval before anything is executed
- On-premise and air-gapped deployment keeps audio inside the customer's own network
- Visibly active development, with a public changelog and several releases in January 2026
Cons
- No mobile application: the tool is desktop only
- No public API and no developer documentation
- No mention of the GDPR anywhere on the site, no data processing agreement and no subprocessor list
- No hosting country or region published for the cloud version
- The terms of service, dated 26 January 2026, still describe the product as free with paid plans yet to come
- Agent Mode covers eight consumer applications; GitHub, Jira, Slack and Notion are not among them
- Support runs through a single generic email address, and no postal address is published
Pricing & Plans
A permanent free plan is available and requires no credit card. The cheapest paid entry point displayed is the Team plan at USD 8.00 per seat per month; the individual Standard plan is listed at USD 10.00 per month. Annual billing is offered at a discount, advertised as 15% on the homepage and 20% on the business and developer pages, although no discounted amount is displayed. Enterprise and on-premise deployments are quoted individually, with no seat minimum. Payments are processed by Paddle, and a full refund may be requested within 14 days.
- limited dictation
- formatting
- connectors and AI assistant
- Terms Dictionary included
- no credit card required
- unlimited dictation
- formatting
- connectors and AI assistant
- subject to anti-abuse guardrails
- advertised as locked for life for the first 5
- 000 paying users
- shared formatting styles
- shared team dictionary
- priority support and centralised billing
- access goes through a call or a contact form
- on-premise or air-gapped deployment
- negotiated contract terms
- no seat minimum
Data, GDPR & hosting
A consolidated view of how NovaVoice handles your data.
GDPR overview
There is no GDPR implementation to report. Across the nine pages collected, in the visible text and in the archived HTML alike, the words GDPR, CCPA, data processing agreement and subprocessor do not appear once. This is silence rather than a refusal: the site neither claims compliance nor denies it. The privacy policy, last updated on 22 November 2025, names no legal basis, no data subject rights, no supervisory authority, no Article 27 representative and no data protection officer, and describes international transfers only as subject to adequate controls. The publisher, Xware VC LLP, gives Astana, Kazakhstan as its country, and the terms are governed by Kazakh law. Those terms do grant a 14-day consumer right of withdrawal, worded in a European style, but cite no regulation. Privacy questions go to one generic address, hello@novavoice.app.
Who owns the data?
The publisher, Xware VC LLP, is the controller named in the privacy policy. It collects your email address, your name and usage data, and states that it may share them with service providers, corporate affiliates, business partners, other users in public areas of the service, and any acquirer in a merger or asset sale. Neither the terms nor the privacy policy say who owns dictated text or captured audio: the question is simply not addressed. You may view, correct or delete your personal data from the account settings or by writing to hello@novavoice.app, though the company reserves the right to keep records where a legal obligation applies.
Reuse rights
The terms place no restriction on what you do with the text NovaVoice produces for you: dictated and reformatted output is inserted straight into your own applications, and no licence, attribution or approval clause covers it. What the terms do prohibit is reselling access, sharing account credentials, using the tool to power third-party services, and extracting data programmatically. On the vendor side, the privacy policy allows personal data to be used to run and improve the service, manage accounts and billing, contact you, send marketing about similar products (which you may decline), and analyse usage. Training AI models on customer content is never mentioned, in either direction, and no opt-out is documented.
Data retention & training
Hosting summary
NovaVoice publishes no hosting country and no hosting region for its cloud service. The only clear statement is a comparative one: the cloud version processes audio on NovaVoice infrastructure, while the on-premise version runs the processing server inside the customer's own data centre, private cloud or air-gapped network, where, in the vendor's words, audio never leaves the network. The privacy policy mentions international transfers only in generic terms, promising adequate controls without naming a mechanism or a jurisdiction. No hosting certification such as SOC 2 or ISO 27001 is claimed anywhere; HIPAA appears once, as a regulation that may apply to a customer, never as a certification held by the vendor. For information, the marketing site itself resolves to an IP address geolocated in Moscow, Russia, on AS9123 JSC TIMEWEB. That describes where the website is served from, not where customer audio is processed, on which the vendor says nothing.
Things to keep in mind
Risks and trade-offs to weigh before adopting NovaVoice.
- An assistant that types for you also listens by design; a hotkey-triggered microphone in an open-plan office or a shared home will pick up what was never meant for it
- Agent Mode sends emails and messages on your behalf, and the approval step is the only thing standing between a misheard sentence and a message a colleague actually receives; approving on autopilot is precisely the habit that forms
- People speak more candidly than they write, so client names, figures and half-formed opinions get dictated that would never be typed; in the cloud version that content leaves your machine, and no hosting location is published for it
- The site never mentions the GDPR, publishes no data processing agreement and no subprocessor list, which leaves a European or regulated buyer with no documentation to rely on
- The privacy policy says nothing about audio, voice, transcription or model training, so the homepage promise that voice data is never stored without consent has no contractual backing
- The Terms Dictionary is designed to hold addresses, loyalty numbers and personal details for one-command insertion, which concentrates sensitive data inside a convenience feature
- Dictating instead of writing removes the pause in which you reread and reconsider; the speed gains advertised are real, but so is the loss of second thoughts before something is sent
Setup & Integrations
Technical difficulty
Straightforward for an individual: download the build for your platform, install it, sign in, and learn one global hotkey. Nothing has to be configured per application, and there is no IDE plugin to install. Agent Mode asks a little more, since each third-party account must be connected, with OAuth cited as the method. Note that the AppImage download button currently points nowhere; only the macOS, Windows and .deb builds have working links. On-premise deployment is a different exercise altogether, an IT project run by the customer's own team on their own servers.
Deployment
Integrations
Behind NovaVoice
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement NovaVoice.
Frequently asked questions
Is there a free version of NovaVoice?
Which devices does NovaVoice run on?
Which applications can Agent Mode control?
How much does NovaVoice cost?
Can the audio stay inside my own network?
Is NovaVoice GDPR compliant?
Is my voice data stored?
Can I dictate source code with it?
How do I contact the publisher?
Can I get a refund?
Should you pick NovaVoice?
NovaVoice is a working product rather than a promise: three downloadable desktop builds, a public changelog that ran from version 1.8.9 to 1.8.21 in a fortnight, published prices, and five comparison pages naming its competitors. What it does well is unusual enough to matter. A Linux desktop client is rare in commercial dictation. Operating at system level means it works inside the applications you already use rather than in a window of its own. And Agent Mode, which is where voice tools usually become uncomfortable, is gated behind an explicit approval step for every single action.
The weakness is not the software, it is the paperwork around it. The site does not mention the GDPR once, offers no data processing agreement, publishes no subprocessor list, and states no hosting location for its cloud version. The privacy policy never mentions audio, voice or transcription, which sits awkwardly next to a homepage FAQ promising that voice data is never stored without consent. The terms of service, updated in January 2026, still describe the tool as free with paid plans yet to arrive. No postal address is published, the publisher Xware VC LLP is hard to trace beyond a country name, and the domain is barely a year old.
The practical reading is straightforward. For an individual, the free plan costs nothing to test and the paid tiers are inexpensive, so the risk is limited to your own time. For a regulated organisation, the on-premise option is the sensible route, because it removes the questions the documentation leaves open. Anywhere in between, obtain written answers on hosting, retention and model training before standardising a team on it.
- Choosing a selection results in a full page refresh.
- Opens in a new window.